diff options
Diffstat (limited to 'o3d/samples/shaders/yuv2rgb-glsl.shader')
-rw-r--r-- | o3d/samples/shaders/yuv2rgb-glsl.shader | 230 |
1 files changed, 230 insertions, 0 deletions
diff --git a/o3d/samples/shaders/yuv2rgb-glsl.shader b/o3d/samples/shaders/yuv2rgb-glsl.shader new file mode 100644 index 0000000..8efcdb6 --- /dev/null +++ b/o3d/samples/shaders/yuv2rgb-glsl.shader @@ -0,0 +1,230 @@ +/* + * Copyright 2009, Google Inc. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are + * met: + * + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following disclaimer + * in the documentation and/or other materials provided with the + * distribution. + * * Neither the name of Google Inc. nor the names of its + * contributors may be used to endorse or promote products derived from + * this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT + * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE + * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +// This shader takes a Y'UV420p image as a single greyscale plane, and +// converts it to RGB by sampling the correct parts of the image, and +// by converting the colorspace to RGB on the fly. + +// Projection matrix for the camera. +uniform mat4 worldViewProjection; + +// These are the input/output parameters for our vertex shader +attribute vec4 position; +attribute vec2 texCoord0; + +// These are the input/output parameters for our pixel shader. +varying vec2 v_texcoord; + +/** + * The vertex shader does nothing but returns the position of the + * vertex using the world view projection matrix. + */ +void main() { + gl_Position = worldViewProjection * position; + v_texcoord = texCoord0; +} + +// #o3d SplitMarker + +// These represent the image dimensions of the SOURCE IMAGE (not the +// Y'UV420p image). This is the same as the dimensions of the Y' +// portion of the Y'UV420p image. They are set from JavaScript. +uniform float imageWidth; +uniform float imageHeight; + +// This is the texture sampler where the greyscale Y'UV420p image is +// accessed. +uniform sampler2D textureSampler; + +varying vec2 v_texcoord; + +/** + * This fetches an individual Y pixel from the image, given the current + * texture coordinates (which range from 0 to 1 on the source texture + * image). They are mapped to the portion of the image that contains + * the Y component. + * + * @param position This is the position of the main image that we're + * trying to render, in parametric coordinates. + */ +float getYPixel(vec2 position) { + position.y = (position.y * 2.0 / 3.0) + (1.0 / 3.0); + return texture2D(textureSampler, position).x; +} + +/** + * This does the crazy work of calculating the planar position (the + * position in the byte stream of the image) of the U or V pixel, and + * then converting that back to x and y coordinates, so that we can + * account for the fact that V is appended to U in the image, and the + * complications that causes (see below for a diagram). + * + * @param position This is the position of the main image that we're + * trying to render, in pixels. + * + * @param planarOffset This is an offset to add to the planar address + * we calculate so that we can find the U image after the V + * image. + */ +vec2 mapCommon(vec2 position, float planarOffset) { + planarOffset += (imageWidth * floor(position.y / 2.0)) / 2.0 + + floor((imageWidth - 1.0 - position.x) / 2.0); + float x = floor(imageWidth - 1.0 - floor(mod(planarOffset, imageWidth))); + float y = floor(floor(planarOffset / imageWidth)); + return vec2((x + 0.5) / imageWidth, (y + 0.5) / (1.5 * imageHeight)); +} + +/** + * This is a helper function for mapping pixel locations to a texture + * coordinate for the U plane. + * + * @param position This is the position of the main image that we're + * trying to render, in pixels. + */ +vec2 mapU(vec2 position) { + float planarOffset = (imageWidth * imageHeight) / 4.0; + return mapCommon(position, planarOffset); +} + +/** + * This is a helper function for mapping pixel locations to a texture + * coordinate for the V plane. + * + * @param position This is the position of the main image that we're + * trying to render, in pixels. + */ +vec2 mapV(vec2 position) { + return mapCommon(position, 0.0); +} + +/** + * Given the texture coordinates, our pixel shader grabs the right + * value from each channel of the source image, converts it from Y'UV + * to RGB, and returns the result. + * + * Each U and V pixel provides color information for a 2x2 block of Y + * pixels. The U and V planes are just appended to the Y image. + * + * For images that have a height divisible by 4, things work out nicely. + * For images that are merely divisible by 2, it's not so nice + * (and YUV420 doesn't work for image sizes not divisible by 2). + * + * Here is a 6x6 image, with the layout of the planes of U and V. + * Notice that the V plane starts halfway through the last scanline + * that has U on it. + * + * 1 +---+---+---+---+---+---+ + * | Y | Y | Y | Y | Y | Y | + * +---+---+---+---+---+---+ + * | Y | Y | Y | Y | Y | Y | + * +---+---+---+---+---+---+ + * | Y | Y | Y | Y | Y | Y | + * +---+---+---+---+---+---+ + * | Y | Y | Y | Y | Y | Y | + * +---+---+---+---+---+---+ + * | Y | Y | Y | Y | Y | Y | + * +---+---+---+---+---+---+ + * | Y | Y | Y | Y | Y | Y | + * .3 +---+---+---+---+---+---+ + * | U | U | U | U | U | U | + * +---+---+---+---+---+---+ + * | U | U | U | V | V | V | + * +---+---+---+---+---+---+ + * | V | V | V | V | V | V | + * 0 +---+---+---+---+---+---+ + * 0 1 + * + * Here is a 4x4 image, where the U and V planes are nicely split into + * separable blocks. + * + * 1 +---+---+---+---+ + * | Y | Y | Y | Y | + * +---+---+---+---+ + * | Y | Y | Y | Y | + * +---+---+---+---+ + * | Y | Y | Y | Y | + * +---+---+---+---+ + * | Y | Y | Y | Y | + * .3 +---+---+---+---+ + * | U | U | U | U | + * +---+---+---+---+ + * | V | V | V | V | + * 0 +---+---+---+---+ + * 0 1 + * + */ +void main() { + // Calculate what image pixel we're on, since we have to calculate + // the location in the image stream, using floor in several places + // which makes it hard to use parametric coordinates. + vec2 pixelPosition = vec2(floor(imageWidth * v_texcoord.x), + floor(imageHeight * v_texcoord.y)); + pixelPosition -= vec2(0.5, 0.5); + // We can use the parametric coordinates to get the Y channel, since it's + // a relatively normal image. + float yChannel = getYPixel(v_texcoord); + + // As noted above, the U and V planes are smashed onto the end of + // the image in an odd way (in our 2D texture mapping, at least), so + // these mapping functions take care of that oddness. + float uChannel = texture2D(textureSampler, mapU(pixelPosition)).x; + float vChannel = texture2D(textureSampler, mapV(pixelPosition)).x; + + // This does the colorspace conversion from Y'UV to RGB as a matrix + // multiply. It also does the offset of the U and V channels from + // [0,1] to [-.5,.5] as part of the transform. + vec4 channels = vec4(yChannel, uChannel, vChannel, 1.0); + + mat4 conversion = mat4(1.0, 0.0, 1.402, -0.701, + 1.0, -0.344, -0.714, 0.529, + 1.0, 1.772, 0.0, -0.886, + 0, 0, 0, 0); + vec3 rgb = (channels * conversion).xyz; + + // This is another Y'UV transform that can be used, but it doesn't + // accurately transform my source image. Your images may well fare + // better with it, however, considering they come from a different + // source, and because I'm not sure that my original was converted + // to Y'UV420p with the same RGB->YUV (or YCrCb) conversion as + // yours. + // + // vec4 channels = vec4(yChannel, uChannel, vChannel, 1.0); + // float3x4 conversion = float3x4(1.0, 0.0, 1.13983, -0.569915, + // 1.0, -0.39465, -0.58060, 0.487625, + // 1.0, 2.03211, 0.0, -1.016055); + // float3 rgb = mul(conversion, channels); + + // Note: The output cannot fully replicate the original image. This is partly + // because WebGL has limited NPOT (non-power-of-two) texture support and also + // due to sRGB color conversions that occur in WebGL but not in the plugin. + gl_FragColor = vec4(rgb, 1.0); +} +// #o3d MatrixLoadOrder RowMajor
\ No newline at end of file |